TikTok + GPT-4: turn comments into content questions
You know the feeling: you’re sure your audience is telling you what they want, but it’s buried in hundreds of TikTok comments across dozens of videos. So you scroll. You copy. You paste. And somehow you still end up guessing what to post next.
This TikTok comment automation hits content creators first, but social media strategists and marketing leads feel it too. Instead of spending about 2 hours hunting for “real questions,” you get a clean list of audience content questions (with source links) you can turn into scripts, hooks, and FAQs.
Below, you’ll see exactly how the workflow runs in n8n, what you need to make it work, and what kind of results to expect once it’s running.
How This Automation Works
The full n8n workflow, from trigger to final output:
n8n Workflow Template: TikTok + GPT-4: turn comments into content questions
flowchart LR
subgraph sg0["Receive Keyword Input Flow"]
direction LR
n0@{ icon: "mdi:cog", form: "rounded", label: "No Operation, do nothing", pos: "b", h: 48 }
n1["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/form.svg' width='40' height='40' /></div><br/>Receive Keyword Input"]
n2["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/httprequest.dark.svg' width='40' height='40' /></div><br/>Search TikTok Users (Dumplin.."]
n3@{ icon: "mdi:swap-vertical", form: "rounded", label: "Split Search Results into Us..", pos: "b", h: 48 }
n4@{ icon: "mdi:cog", form: "rounded", label: "Limit to 3 Users (Optional)", pos: "b", h: 48 }
n5@{ icon: "mdi:swap-vertical", form: "rounded", label: "Loop Through TikTok Users", pos: "b", h: 48 }
n6@{ icon: "mdi:cog", form: "rounded", label: "Wait to Respect Rate Limits", pos: "b", h: 48 }
n7["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/httprequest.dark.svg' width='40' height='40' /></div><br/>Get TikTok Profile Videos (D.."]
n8@{ icon: "mdi:swap-vertical", form: "rounded", label: "Split Videos from Profile", pos: "b", h: 48 }
n9@{ icon: "mdi:swap-vertical", form: "rounded", label: "Loop Through Videos", pos: "b", h: 48 }
n10["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/httprequest.dark.svg' width='40' height='40' /></div><br/>Get Comments for Each Video"]
n11["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Extract Clean Comments (Pyth.."]
n12@{ icon: "mdi:robot", form: "rounded", label: "Find Top Viewer Questions (G..", pos: "b", h: 48 }
n13@{ icon: "mdi:cog", form: "rounded", label: "Insert Result into DataTable", pos: "b", h: 48 }
n9 --> n0
n9 --> n10
n1 --> n2
n5 --> n8
n5 --> n6
n8 --> n9
n10 --> n11
n4 --> n5
n6 --> n7
n13 --> n9
n11 --> n12
n3 --> n4
n12 --> n13
n2 --> n3
n7 --> n5
end
%% Styling
classDef trigger fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
classDef ai fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
classDef aiModel fill:#e8eaf6,stroke:#3f51b5,stroke-width:2px
classDef decision fill:#fff8e1,stroke:#f9a825,stroke-width:2px
classDef database fill:#fce4ec,stroke:#c2185b,stroke-width:2px
classDef api fill:#fff3e0,stroke:#e65100,stroke-width:2px
classDef code fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
classDef disabled stroke-dasharray: 5 5,opacity: 0.5
class n1 trigger
class n12 ai
class n2,n7,n10 api
class n11 code
classDef customIcon fill:none,stroke:none
class n1,n2,n7,n10,n11 customIcon
The Problem: TikTok comments are gold, but messy
Comments are where the truth is. People ask what they’re confused about, what they want next, what they tried, what failed, and what they’re embarrassed to admit. The problem is that TikTok doesn’t hand you those insights in a usable format. You have to open profiles, check recent videos, dig through threads, and then manually translate messy text into “questions worth answering.” Miss a few days and you’re back at zero, trying to remember what you saw last week.
It adds up fast. Here’s where it usually breaks down in the real world.
- You end up sampling a handful of comments, which means you miss the repeated patterns that actually drive good content.
- Copy-pasting into a doc is slow, and it’s easy to lose the video link you’ll want later for context.
- Team workflows get awkward because “comment research” lives in someone’s browser tabs, not in a shared planning system.
- When you finally sit down to plan content, you still have to rewrite everything into clear, publishable questions.
The Solution: Turn a keyword into audience questions (automatically)
This workflow takes one simple input (a keyword) and turns it into a practical content research deliverable: a short list of recurring audience questions, grouped and cleaned, each with the source video link. It starts by searching TikTok for creator profiles related to your keyword using Dumpling AI. Then it pulls the latest videos from those profiles, collects the comments, and runs a quick cleaning step so the text is readable and consistent. Finally, GPT-4 reviews the cleaned comments and extracts the top questions your audience keeps asking. The workflow stores the results in an n8n Data Table so you can review, sort, and plan content without digging through TikTok again.
The workflow begins when you submit a keyword through a form trigger. After that, it cycles through creators and videos in batches, pauses briefly to avoid rate limits, and keeps looping until it has processed the newest content. When the AI finishes, you get a neat “question library” with links you can reference later.
What You Get: Automation vs. Results
| What This Workflow Automates | Results You’ll Get |
|---|---|
|
|
Example: What This Looks Like
Say you’re building content around one keyword and you want ideas from 10 creators. Manually, you might spend about 10 minutes per creator just to find recent videos, then another 10 minutes per video digging through comments and saving links. Even if you only check 2 videos each, that’s roughly 3–4 hours of research. With this workflow, you submit the keyword in under a minute, let the batches run (often about 20–30 minutes depending on volume), and you get a ready-to-use question list with video links waiting in your table.
What You’ll Need
- n8n instance (try n8n Cloud free)
- Self-hosting option if you prefer (Hostinger works well)
- Dumpling AI for TikTok search and data access.
- OpenAI (GPT-4) to group comments into clear questions.
- Dumpling AI API key (get it from your Dumpling AI dashboard).
Skill level: Intermediate. You’ll connect credentials, review a prompt, and confirm your Data Table fields.
Don’t want to set this up yourself? Talk to an automation expert (free 15-minute consultation).
How It Works
You submit a keyword through a form trigger. That keyword becomes the “research topic” for the run, so you can reuse the workflow for any niche or campaign theme.
TikTok creators and videos are pulled in batches. The workflow calls Dumpling AI via HTTP requests, splits results into manageable chunks, and caps the number of users so you don’t accidentally fetch a huge dataset on day one.
Comments are collected, then cleaned before AI sees them. A code step (Python in n8n) removes junk text and normalizes the comment content, which honestly makes a big difference in question quality.
GPT-4 extracts recurring questions and the workflow stores them. The AI step outputs the “top questions,” and the Data Table stores the questions alongside useful metadata like video links so you can review and plan content quickly.
You can easily modify the keyword source to come from Telegram or a schedule based on your needs. See the full implementation guide below for customization options.
Step-by-Step Implementation Guide
Step 1: Configure the Form Trigger
This workflow starts when a user submits keywords via a form, which then drives the TikTok search requests.
- Add and open Keyword Form Intake.
- Set Form Title to
Tik Tok Search. - Confirm a form field labeled Keywords is present in Form Fields.
- Verify the execution flow starts from Keyword Form Intake to Query TikTok Users API.
Step 2: Connect TikTok Search Requests
These nodes call the TikTok user search and profile video APIs and prepare results for batch processing.
- Open Query TikTok Users API and set URL to
https://app.dumplingai.com/api/v1/search-tiktok-users. - Set Method to
POSTand ensure Send Body is enabled. - In Body Parameters, set query to
{{ $json.Keywords }}. - Credential Required: Connect your
httpHeaderAuthcredentials in Query TikTok Users API. - Open Fetch Profile Videos API and set URL to
https://app.dumplingai.com/api/v1/get-tiktok-profile-videoswith MethodPOST. - Set Body Parameters handle to
{{ $json.user.search_user_name }}. - Credential Required: Connect your
httpHeaderAuthcredentials in Fetch Profile Videos API.
query and handle), the DumplingAI endpoints will return empty results.Step 3: Configure Batching and Rate Limiting
Batching and pauses keep the workflow within API limits while iterating over users and videos.
- In Divide User Results, set Field to Split Out to
userList. - In Cap User Count, set a limit if you want to restrict how many users are processed per run.
- Open Iterate User Batches to batch through users; keep default options unless you need specific batch sizing.
- Configure Rate Limit Pause with Amount set to
15to throttle API calls. - In Separate Video List, set Field to Split Out to
aweme_list. - Use Cycle Video Batches to iterate through videos; it outputs to Retrieve Video Comments and loops back via Store Results Table per execution flow.
Step 4: Extract and Prepare Comment Data
This stage pulls comments for each video and cleans them into a structured list for AI analysis.
- Open Retrieve Video Comments and set URL to
https://app.dumplingai.com/api/v1/get-tiktok-video-commentswith MethodPOST. - Set Body Parameters url to
{{ $json.share_info.share_url }}. - Credential Required: Connect your
httpHeaderAuthcredentials in Retrieve Video Comments. - In Clean Comment Text, set Language to
pythonand paste the provided code to extractCommententries. - Confirm execution flow from Retrieve Video Comments → Clean Comment Text → Derive Top Questions.
Step 5: Set Up AI Question Extraction
The AI node analyzes comments to produce a ranked list of commonly asked questions.
- Open Derive Top Questions and select model
gpt-4.1. - Confirm the system message instructs the model to identify top recurring questions.
- Set the user message to include
Here is the comment data: {{ JSON.stringify($json.Comment) }}. - Credential Required: Connect your
openAiApicredentials in Derive Top Questions.
Comment array.Step 6: Configure Output Storage
Results are written to a data table with the video URL and extracted questions.
- Open Store Results Table and select the target data table ID (for example,
Tik Tok Keywords). - Set the Videos column value to
{{ $('Cycle Video Batches').item.json.share_info.share_url }}. - Set the Keywords column value to
{{ $json.message.content }}. - Confirm Store Results Table connects back to Cycle Video Batches to continue processing.
Step 7: Test and Activate Your Workflow
Validate the workflow end-to-end and then turn it on for production use.
- Click Execute Workflow and submit the form in Keyword Form Intake with a sample keyword.
- Verify that Query TikTok Users API returns results and that batches pass through Iterate User Batches and Cycle Video Batches.
- Check that Derive Top Questions outputs a ranked list of questions in message.content.
- Confirm rows appear in Store Results Table with video URLs and extracted questions.
- Toggle the workflow Active to enable ongoing use.
Common Gotchas
- Dumpling AI credentials can expire or need specific permissions. If things break, check your Dumpling AI dashboard and the n8n credential test first.
- If you’re using Wait nodes or external fetching, processing times vary. Bump up the wait duration if downstream requests fail because the previous batch hasn’t returned results yet.
- Default prompts in the GPT-4 node are generic. Add your brand voice early or you will keep rewriting “pretty good” questions into usable ones.
Frequently Asked Questions
About 30 minutes if your API keys are ready.
No. You’ll mostly connect accounts and adjust a couple of fields and prompts.
Yes. n8n has a free self-hosted option and a free trial on n8n Cloud. Cloud plans start at $20/month for higher volume. You’ll also need to factor in OpenAI API usage (usually a few dollars a month for light research) and whatever Dumpling AI charges on your plan.
Two options: n8n Cloud (managed, easiest setup) or self-hosting on a VPS. For self-hosting, Hostinger VPS is affordable and handles n8n well. Self-hosting gives you unlimited executions but requires basic server management.
Yes, and it’s a common upgrade. You can swap the keyword form trigger for a scheduled trigger, then feed in a fixed list of keywords (or pull them from Google Sheets). Many teams also tweak the GPT-4 prompt to output a specific format like “hook + question + angle,” which makes content production even faster. If you route results into Google Sheets or Excel 365 instead of the n8n Data Table, your whole team can work from one shared backlog.
Usually it’s an expired API key or the key is stored in the wrong n8n credential entry. Regenerate the key in Dumpling AI, update it in n8n, then re-run the workflow from the HTTP request step. If it fails only on larger runs, you may be hitting rate limits, so increase the Wait node duration and reduce how many creators you process per batch.
On n8n Cloud Starter, you can run a few thousand executions per month, which is plenty for weekly research. If you self-host, there’s no execution limit, and scale mostly depends on your server and API rate limits. Practically, most teams start with 5–20 creators per keyword and expand once the output quality looks good.
Often, yes, because this workflow needs batching, waiting, looping, and a bit of cleaning before the AI step. n8n handles those patterns cleanly, and self-hosting keeps your run volume from getting expensive. Zapier or Make can be fine for simple “new thing → send thing” automations, but this one is closer to a mini data pipeline. If your output needs to be reliable (and not randomly cut off mid-run), n8n is usually the calmer choice. Talk to an automation expert if you’re not sure which fits.
Once this is running, your content planning stops starting from scratch. The workflow handles the digging, and you get to focus on shipping the next good post.
Need Help Setting This Up?
Our automation experts can build and customize this workflow for your specific needs. Free 15-minute consultation—no commitment required.
